Skip to main content

AI Writing Methodology: Cursor + Voice Transcription Workshop

July 12, 2025 - Gary Sheng's Gauntlet AI Workshop​

Overview​

This document outlines the AI-assisted writing methodology demonstrated by Gary Sheng during his workshop at Gauntlet AI. The approach emphasizes context engineering over prompt engineering, using voice transcription and Cursor's AI capabilities to create high-quality content efficiently.

Core Philosophy: Context Engineering Over Prompt Engineering​

The Fundamental Shift​

"There's basically, in my view, only one way, one approach to writing with AI that makes any sense. I'm sure you've heard of the distinction between, or at least the clarification of, from prompt engineering to context engineering."

Essential Context Elements​

  1. Audience Understanding - Who will read this?
  2. Tone Specification - What voice/style should it have?
  3. Relevant Literature - What existing work informs this?
  4. Primary Evidence - What specific examples/data support the points?
  5. Past Writing Samples - Examples of your best work for style reference

The Voice-to-AI Writing Process​

Step 1: Voice Capture​

  • Tool: Apple Voice Recorder (auto-transcribes)
  • Method: Record 5-30 minutes of natural conversation about the topic
  • Approach: Talk naturally as if explaining to a colleague or friend
  • Key Quote: "What I'll do, like, if I have something that is from my heart that I want to express, now I've gotten pretty good at, like, just talking to myself. Or, if I really need to, I'll put on grok voice mode and just talk in grok voice mode for like 30 minutes."

Step 2: Context Setup in Cursor​

  • Tool: Cursor (AI-powered code editor adapted for writing)
  • Method: Create a dedicated writing repository
  • Context Loading: Include transcript, past writing samples, and specific instructions
  • Benefit: "I have a repo where I basically do all my writing, and it just... Every new piece of writing is there, because that's where I'm drafting it and editing it to the point of completion. And so it's always being able to reference all of the writing that I have."

Step 3: AI-Assisted Drafting​

  • Initial Generation: Let AI create first draft from voice transcript
  • Iteration: Provide specific feedback for improvements
  • Refinement: Use Cursor's diff system to accept/reject specific changes

Step 4: Human-AI Collaboration​

  • Feedback Loop: Give detailed, specific feedback about tone, structure, content
  • Technical Advantage: "If you do one comment about how you want it changed, you can see seven diffs. You can accept, reject, right? That level of granularity is very important."
  • Key Principle: "Don't accept the first drafts. Keep drilling at it with manual typing, and also you can just chat with your essays now, right?"

Time Efficiency Framework​

The 15-Minute Formula​

  • 5 minutes: Voice conversation/interrogation
  • 10 minutes: AI-assisted editing and refinement
  • Result: Content that would otherwise take "infinite time or never happen"

Comparison to Traditional Methods​

  • Traditional: Often results in writer's block or abandoned projects
  • AI-Assisted: Transforms scattered thoughts into structured content rapidly
  • Key Insight: "What would probably take either infinite time or it would never happen because you're like, oh, I hate writing. It's now 15 minutes, right?"

Technical Implementation​

Required Tools​

  1. Voice Recorder: Apple Voice Recorder (auto-transcription) or similar
  2. AI Editor: Cursor (preferred) or similar AI-powered writing environment
  3. Context Repository: Dedicated folder/repo for all writing projects

Setup Process​

  1. Create dedicated writing repository in Cursor
  2. Organize past writing samples by type/quality
  3. Develop template for context setting
  4. Practice voice-to-transcript workflow

Best Practices​

  • Conciseness: Explicitly instruct AI to be concise and avoid repetition
  • Specificity: Provide detailed feedback rather than generic requests
  • Iteration: Expect multiple rounds of refinement
  • Context Accumulation: Each new piece adds to your writing context library

Content Strategy Integration​

Voice Authenticity​

  • Principle: Start with authentic voice through natural speech
  • Benefit: Maintains personal tone and genuine expression
  • Quote: "AI often is very verbose. So I often say, you can... There's a lot of nonsense transcripts. You can just cancel it. But often you have to literally say, be concise."

Audience-Specific Adaptation​

  • Method: Clearly define target audience in context
  • Application: Tailor complexity, examples, and tone accordingly
  • Feedback: Test with intended audience before publishing

Community Building Application​

  • Framework: Document learning journey for others following similar paths
  • Strategy: Share insights about daily challenges and discoveries
  • Quote: "The easiest formula with us going through Gauntlet and becoming AI engineers is reflect on what was the most challenging thing that you went through today and what was your discovery pattern of that. And then try and just relay that back in as elementary terms as possible."

Limitations and Considerations​

What AI Does Well​

  • Long-form content: Excellent for blog posts, articles, detailed explanations
  • Structure: Good at organizing scattered thoughts into coherent flow
  • Editing: Effective at refining tone, clarity, and style

What AI Struggles With​

  • Pithy tweets: "From my experience, AI sucks at those Naval one-liner type things"
  • Authentic voice: Without proper context, tends toward generic corporate speak
  • Original insights: Requires human input for unique perspectives and experiences

Human Skills Still Required​

  • Initial insight: The core ideas and experiences must come from human
  • Quality judgment: Ability to assess and refine AI output
  • Authentic expression: Maintaining personal voice throughout the process

Workshop Outcomes​

Demonstrated Effectiveness​

The workshop showed how a 13-minute voice conversation could be transformed into a structured blog post outline and draft within minutes, with iterative refinement producing increasingly polished content.

Participant Feedback Integration​

Yash's real-time feedback during the demo showed how specific critiques ("feels extremely AI-written," "jarring sentence," "doesn't feel connected") led to immediate improvements in subsequent AI iterations.

Community Impact​

The methodology enables individuals to share their learning journey and expertise more effectively, creating compound benefits for both personal brand building and community knowledge sharing.

Key Success Principles​

  1. Context is Everything: The quality of AI output directly correlates with the quality and specificity of context provided
  2. Voice First: Start with authentic verbal expression, then refine with AI assistance
  3. Iterate Relentlessly: First drafts are starting points, not endpoints
  4. Maintain Human Judgment: AI is a tool for efficiency, not replacement for human insight
  5. Build Systematically: Each piece of writing adds to your context library, improving future outputs

Future Applications​

Personal Brand Building​

  • Document expertise development in real-time
  • Share learning insights with authenticity and consistency
  • Build audience through valuable, regular content creation

Community Contribution​

  • Lower barriers to knowledge sharing
  • Enable more people to contribute written insights
  • Create compound learning effects within groups

Professional Development​

  • Demonstrate growing expertise publicly
  • Create portfolio of thought leadership content
  • Build reputation as subject matter expert

This methodology represents a significant shift in how high-quality writing can be produced, making it accessible to those who previously struggled with traditional writing processes while maintaining authenticity and personal voice.